Reviewers' comments:
Reviewer #2 (Remarks to the Author):
This article tackles an interesting question, that of the evolving relationship between human populations and water, by combining information on past censuses and on freshwater source location. The methods are well explained and the paper is easy to read.
Regrettably, the authors "assumed temporal stationarity of the spatial distribution of both major rivers and groundwater over time" (manuscript lines 265-266). This is simply not something that happens, which invalidates all their subsequent work: the quantitative aspects are not believable, nor is an estimate of the error given.
The authors should at least reference some geomorphological studies to support their course of action. However, I believe it will be hard to do so, because river courses and groundwater reservoirs can and do change radically through time. This happens both as a result of internal factors (for example, rivers migrating) and as a response to external forces (for instance, to name just two examples on opposite sides of the duration spectrum, earthquakes and tectonic subsidence). Although I do not work on the subject, I imagine there are many geomorphological studies reconstructing previous courses of North American rivers that could be combined with the same census data used in the manuscript to greatly improve this study.
Thus, my recommendation is REJECT, but I believe a substantial revision which takes into account the histories/movements of the different rivers (and, if at all possible, although I realise this will be harder, groundwater reservoirs) would warrant a RESUBMISSION.
Reviewer #3 (Remarks to the Author):
This is an important paper investigating the history of human settlements near water in the US from 1790 to 2010. It is important because it provides evidence for assertions many make and think are self-evident, and investigates regional differences in patterns. I have a few minor items and two moderate issues of concern.
Minor items: 1. There is an extra comma after the word "source" in the first paragraph after the abstract.
2. It would be helpful to provide a brief explanation of why the 1960 census data are not usable/available.
3. I think a better color ramp is needed for figure 1A. It is difficult to discern between the 3 blue classes and the two green ones.
4. The text for figure 3 should explain what the data points are (DMR classes, I'm assuming, for 3A? The census year for 3C?) And what the colors mean. Is red to blue simply the decades between 1790 and 1870? Why does 3A. progress red to blue but then 3B go blue to red?
5. Similarly the text for figure 4 should explain what the data points are. Also why does 4F only have data since ~1830? (I'm assuming, as a western frontier, that there were no settlements there prior to that date).
6. Again, text for supplementary figures should indicate what the data points are (e.g., I'm assuming S1B are DMR classes again?).
Moderate issues: 1. Is there a relationship between aquifer type and river density? I'm wondering if correlation between these two could confound analysis. A brief treatment in the text would probably suffice.
2. I disagree with the evaluation that "major aquifers with high recharge rates, especially those at recharge rates of 100-300 mm/a, were associated with low desirability of living close to major rivers (Supplementary Fig. 2)." This figure seems to show that *only* the 100-300 mm/a aquifer type shows a low desirability (positive slope), while all the others should a generally negative slope.
I believe all of these issues are fixable and do not detract from the overall importance of this paper.
Elisabeth Larson
[email protected]
Reviewer #4 (Remarks to the Author):
This study addresses the evolution of human population distance to water for the conterminous US over the period of 1790-2010. This is an innovative and important research question in the newly developed discipline of socio-hydrology. The authors have made excellent headway in linking humans and water and provided a quantitative demonstration that human population preferred to live close to major rivers in pre-industrial periods and have moved to areas overlying major aquifers since industrial revolution. This paper makes an exciting contribution to the study of Socio-hydrology by incorporating spatial perspective and capturing long-term (over two centuries) human-water coevolution based on available historical data. The method is well described and based on solid scientific background. The topic of the paper can appeal a wide range of readers who are working on water resources sustainability, and broadly human-nature interaction. The subject matter is therefore suitable for Nature Communications.
Overall, the study is well designed and well written. With appropriate method and innovative results, this is an excellent paper and has very good potential for publication in Nature Communications. However, there are some parts which requires further discussion or elaboration. I recommend publication subject to the minor comments suggested below.
1) This study showed the evolution of human water distance for the conterminous US influenced by societal progress and technological development, including industrial revolution and groundwater pumping technique. However, the uncertainty could come from water diversion projects especially during the recent several decades (maybe half
century). I acknowledge that the authors explained the importance of water diversion. For example, in Line 163, “water supply in post-industrial societies has evolved from “people go to water” to “water goes to people”, and it is understandable that the current knowledge and data availability might limit a further quantitative analysis for water diversion issue over the whole conterminous US. However, it would be helpful to include some water diversion projects and discuss their impact on human water distance here.
2) This paper analyzed the temporal trends of human population distribution in relation to both major rivers and groundwater for the conterminous US. The dynamic trend of the importance of major rivers was nicely captured by the relationship between normalized population density and distance to major rivers as well as their slopes (Figure 3, a-c). The finding with regard to aquifer type is interesting and useful- the advanced groundwater pumping technology around 1940 influenced the human population distribution overlying aquifers. However, the statement on groundwater is somewhat weak, since the turning point of the year of 1940 was currently identified through eyeballing (Figure 3, d). I would like to see a quantitative basis of this to strengthen the statement.
3) The paper presented 3 major typologies of temporal changes in NPD vs DMR and related them to the local settlement history, water resource endowments, and climate condition. This revealed regional heterogeneity and enriched the analysis. However, it is not clear what this result can be used for. I would therefore suggest that the authors add some discussion on the broader impacts or relevance with respect to the result on regional heterogeneity.
4) Line 197: “Ensuring adequate infrastructure investment is therefore extraordinarily essential to sustain stable water supplies and decrease the vulnerability of water supply systems.” Should the vulnerability of water supply systems get increased with increasing investment?
Reviewer #2 (Remarks to the Author): This article tackles an interesting question, that of the evolving relationship between human populations and water, by combining information on past censuses and on freshwater source location. The methods are well explained and the paper is easy to read. Regrettably, the authors "assumed temporal stationarity of the spatial distribution of both major rivers and groundwater over time" (manuscript lines 265-266). This is simply not something that happens, which invalidates all their subsequent work: the quantitative aspects are not believable, nor is an estimate of the error given. The authors should at least reference some geomorphological studies to support their course of action. However, I believe it will be hard to do so, because river courses and groundwater reservoirs can and do change radically through time. This happens both as a result of internal factors (for example, rivers migrating) and as a response to external forces (for instance, to name just two examples on opposite sides of the duration spectrum, earthquakes and tectonic subsidence). Although I do not work on the subject, I imagine there are many geomorphological studies reconstructing previous courses of North American rivers that could be combined with the same census data used in the manuscript to greatly improve this study. Thus, my recommendation is REJECT, but I believe a substantial revision which takes into account the histories/movements of the different rivers (and, if at all possible, although I realise this will be harder, groundwater reservoirs) would warrant a RESUBMISSION. RESPONSE: We appreciate the reviewer’s suggestion to explicitly consider river course changes. We have incorporated a discussion (lines 211-213, 216-223) with appropriate findings from the river geomorphology literature and including data from global databases of river widths and reservoir construction. Specifically, 1) River course migration: We have considered the drivers of river course migration and compared published migration rates, with the overall conclusion that migration widths are much smaller than the resolution of our analysis (1 km). Thus, at this spatial scale the assumption of stationarity of river networks is reasonable. Please see Lines 211-213 and the Supplementary Note for details. 2) River widths: Despite the generally minimal effect from river migration, we agree with the reviewer about the importance of the coverage of river courses. We have now included river width data from the Global River Widths from Landsat (GRWL) database, with approximately 7500 km of rivers with width larger than 1 km. This resulted in a small increase in the areas (0.8%) with distance to major rivers (DMR) of zero, compared to our previous result assuming a constant river width of 1 km. (Lines 281-288) 3) Reservoirs: Again, inspired by the reviewer’s comments, we now include the temporal evolution of reservoirs from the Global Reservoir and Dam (GRanD) dataset to identify reservoirs based of the decade of construction. We use the reservoir areas from GRanD and again this resulted in a small increase (2.6%) in Page | 1
the areas with DMR of zero, compared to the reference river width based on GRWL data. Thus our calculated DMR for each decade now includes geomorphological alteration over time. (Lines 292-295) 4) Aquifers: The timescales for large-scale changes in aquifer boundary migration are geological rather than human in scale. However, the importance of aquifer volume changes due to groundwater pumping has been growing in recent decades, and is expected to continue to grow in significance in coming decades. We have added a discussion in lines 216-223. After including river width and considering the growth of reservoirs, we re-calculated all of our results and made the associated changes in the text, Figures, and tables. However, the general results were still the same and our conclusions on human distance to water remained. See lines 105-111, 147, 154, 161, and 165-166 for the detailed modifications. Reviewer #3 (Remarks to the Author): This is an important paper investigating the history of human settlements near water in the US from 1790 to 2010. It is important because it provides evidence for assertions many make and think are self-evident, and investigates regional differences in patterns. I have a few minor items and two moderate issues of concern. Minor items: 1. There is an extra comma after the word "source" in the first paragraph after the abstract. RESPONSE: We appreciate very much that the reviewer scrutinized this and have removed the extra comma in Line 37. 2. It would be helpful to provide a brief explanation of why the 1960 census data are not usable/available. RESPONSE: We agree with the reviewer and have added a brief explanation (lines 248-251): “Note our analysis excluded 1960 because these data are still not available due to the unusually restrictive data suppression strategy employed by the 1960 Census (https://www.nhgis.org/user-resources/faq#1960_Data).” FYI, clipped from the link above:
Page | 2
3. I think a better color ramp is needed for figure 1A. It is difficult to discern between the 3 blue classes and the two green ones. RESPONSE: We agree with the reviewer. Also, the distance classes were hard to discern at the continental scale. We therefore have reduced the number of the distance classes to make their distribution patterns more discernible. The updated color ramp and distance classes are shown in Figure 1A below.
4. The text for figure 3 should explain what the data points are (DMR classes, I'm assuming, for 3A? The census year for 3C?) And what the colors mean. Is red to blue simply the decades between 1790 and 1870? Why does 3A. progress red to blue but then 3B go blue to red? RESPONSE: As suggested by the reviewer, we have added an explanation to the caption that the data points for Figure 3a and 3b are based on DMR classes while those for Figure 3c and 3d are based on census year. These are also labeled on the X axes. Also, as suggested by the reviewer, we have added an explanation to the caption regarding the color gradient. In Figure 3a we use the red-to-blue gradient to indicate Page | 3
the trend for the decades to 1870 and in Figure 3b the gradient from blue to red to refer to decades from 1870. This choice is to highlight with the darker blues stronger desirability to live close to major rivers (steeper slopes) and the darker red colors relatively weaker desirability (smaller slopes). Also, we changed the color ramp for Figure 3d and increased the symbol size to make them easier to observe.
Figure 3 | The dynamic human population distribution in relation to major rivers and groundwater in the conterminous US. The relationship between normalized population density (NPD) and distance to major rivers (DMR) from a 1790 to 1870 (red to blue), and b 1870 to 2010 (blue to red). c The changing desirability of living close to major rivers, reflected in the slope of NPD vs DMR. d Mean population density overlying each aquifer type from 1790 to 2010. We use the red-to-blue gradient to indicate the trend for the decades to 1870 in a and the gradient from blue to red for decades from 1870 in b. The data points for a-b are based on DMR classes while those for c-d are based on census year. 5. Similarly the text for figure 4 should explain what the data points are. Also why does 4F Page | 4
only have data since ~1830? (I'm assuming, as a western frontier, that there were no settlements there prior to that date). RESPONSE: As suggested by the reviewer we have added further explanation to the caption, similar to the above comment. we previously included slopes with p < 0.05. As the reviewer suggested, we have systematically determined the starting years of analysis for each HUC based on their settlement history and only included data in Figure 4d-f after these initial years. This is now explained in the methods (line 327-331) and in the caption of Figure 4. We have added the initial decade of analysis by HUC in new supplementary Figure 7 and their dynamic settlement area percentages in supplementary Table 4. Lines 327-331: “Only statistically significant slopes (p < 0.05) between NPD and DMR were used. Due to the westward migration of primarily European settlers, the settled portions of each HUC expanded gradually over time. The initial decade of analysis was determined for each HUC based on the decade when the settled area reached 90% of the total area (Supplementary Table 4 and Supplementary Fig. 7).”
Figure 4 | Three major types of trajectories of human water coevolution, based on the relationship between normalized population density (NPD) and distance to major rivers (DMR), with HUC 1, 3, and 15 as examples. a HUC 1, 1850 to 2010, b HUC 3, 1850 to 2010, Page | 5
and c HUC 15, 1870 to 2010. d-f The dynamics of the slope of NPD vs DMR for HUC 1, 3, and 15, 1790-2010. Only slopes with p < 0.05 during the settlement periods are shown. The data points for a-c are based on DMR classes while those for d-f are based on census year.
Page | 6
Supplementary Table 4. The settlement area percentage for each HUC in the conterminous US, from 1790 to 2010. Decade
HUC1
HUC2
HUC3
HUC4
HUC5
HUC6
HUC7
HUC8
HUC9
HUC10
HUC11
HUC12
HUC13
HUC14
HUC15
HUC16
HUC17
HUC18
1790
99.9
99.9
38.5
20.8
48.9
13.1
0.0
1.4
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
1800
100.0
99.9
39.1
100.0
97.7
34.9
49.7
5.4
13.5
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
1810
100.0
100.0
44.3
78.7
90.0
59.2
53.7
47.4
13.5
4.0
14.1
0.1
0.0
0.0
0.0
0.0
0.0
0.0
1820
100.0
99.9
63.6
90.3
97.3
82.7
51.8
71.8
13.5
6.7
17.6
1.3
0.0
0.0
0.0
0.0
0.0
0.0
1830
100.0
100.0
87.2
100.0
99.2
89.0
57.0
85.3
13.5
7.2
20.9
1.3
0.0
0.0
0.0
0.0
0.0
0.0
1840
100.0
100.0
100.0
92.3
100.0
100.0
81.9
99.9
33.4
8.0
19.3
1.3
0.0
0.0
0.0
0.0
0.0
0.0
1850
100.0
100.0
100.0
100.0
100.0
100.0
92.0
100.0
98.5
20.2
38.2
96.7
58.4
7.3
2.0
11.6
97.5
96.8
1860
100.0
99.9
100.0
96.6
100.0
100.0
98.8
99.9
46.6
28.2
44.4
73.6
94.3
96.3
100.0
96.2
99.8
100.0
1870
100.0
100.0
100.0
97.1
100.0
100.0
100.0
99.9
89.7
96.8
49.6
84.6
100.0
98.8
94.0
98.8
100.0
100.0
1880
100.0
99.9
100.0
100.0
100.0
100.0
100.0
99.9
80.6
93.2
69.0
94.8
99.8
100.0
100.0
100.0
100.0
100.0
1890
100.0
99.9
100.0
100.0
100.0
100.0
100.0
99.9
99.2
98.1
75.6
98.5
100.0
100.0
100.0
100.0
99.7
100.0
1900
100.0
99.9
100.0
100.0
100.0
100.0
100.0
99.9
100.0
100.0
99.7
100.0
100.0
100.0
100.0
100.0
100.0
100.0
1910
100.0
99.9
100.0
100.0
100.0
100.0
100.0
99.9
100.0
100.0
100.0
100.0
100.0
100.0
100.0
100.0
100.0
100.0
1920
100.0
99.9
100.0
100.0
100.0
100.0
100.0
99.9
100.0
100.0
100.0
100.0
100.0
100.0
100.0
100.0
100.0
100.0
1930
100.0
99.9
100.0
100.0
100.0
100.0
100.0
99.9
100.0
100.0
100.0
100.0
100.0
100.0
100.0
100.0
100.0
100.0
1940
100.0
99.9
100.0
100.0
100.0
100.0
100.0
99.9
100.0
100.0
100.0
100.0
100.0
100.0
100.0
100.0
100.0
100.0
1950
100.0
99.9
100.0
100.0
100.0
100.0
100.0
99.9
100.0
100.0
100.0
100.0
100.0
100.0
100.0
100.0
100.0
100.0
1970
100.0
99.9
100.0
100.0
100.0
100.0
100.0
100.0
100.0
100.0
100.0
100.0
100.0
100.0
100.0
99.9
100.0
100.0
1980
100.0
99.9
100.0
100.0
100.0
100.0
100.0
100.0
100.0
99.9
100.0
100.0
99.9
100.0
100.0
99.9
99.9
100.0
1990
100.0
99.8
100.0
100.0
100.0
100.0
100.0
100.0
100.0
100.0
100.0
100.0
100.0
100.0
100.0
100.0
100.0
100.0
2000
100.0
99.9
100.0
100.0
100.0
100.0
100.0
100.0
100.0
100.0
100.0
100.0
100.0
100.0
100.0
100.0
100.0
100.0
2010
100.0
99.9
100.0
100.0
100.0
100.0
100.0
100.0
100.0
100.0
100.0
100.0
100.0
100.0
100.0
100.0
100.0
100.0
Note. The starting year of analysis was determined for each HUC as the year when the area percentage reached 90%. Page | 7
Supplementary Figure 7. The initial decade of analysis for each HUC.
6. Again, text for supplementary figures should indicate what the data points are (e.g., I'm assuming S1B are DMR classes again?). RESPONSE: As suggested by the reviewer, and similarly to the figures above, we have added explanatory text to the caption that these are the DMR classes. Moderate issues: 1. Is there a relationship between aquifer type and river density? I'm wondering if correlation between these two could confound analysis. A brief treatment in the text would probably suffice. RESPONSE: As suggested by the reviewer, we added additional explanatory text in the methods (lines 316-319) about this analysis that was previously shown in Supplementary Figure 2 and explained in lines 132-140. We combined DMR zones and aquifer types to analyze their interacting effect on population density. “In addition, considering the complex interaction between rivers and aquifers, we calculated the NPD values associated with DMR for each of the 10 aquifer types and analyzed their combined effect on population density using 2010 as a snapshot” 2. I disagree with the evaluation that "major aquifers with high recharge rates, especially those at recharge rates of 100-300 mm/a, were associated with low desirability of living close to major rivers (Supplementary Fig. 2)." This figure seems to show that *only* the 100-300 mm/a aquifer type shows a low desirability (positive slope), while all the others should a generally negative slope.
Page | 8
RESPONSE: We agree with the reviewer, and we have now calculated the desirability of each aquifer type (indicated by the slope of NPD vs DMR) and listed the results in the new Supplementary Table 1. Also, we have modified the main text to clarify this point (Lines 133-140) “Across the US, areas with the relatively low-yielding local and shallow aquifers (aquifer types 33 and 34), were associated with strong desirability of living close to rivers (slope ≤ -0.03, Supplementary Fig. 2 and Table 1), regardless of recharge rate. Conversely, complex and major aquifers with high recharge rates (aquifer types 24, 13, 14, and 15) had relatively lower desirability of living close to rivers (slope ≥ -0.02). Major aquifers with recharge rates of 100-300 mm/a (aquifer type 14) had the lowest desirability of living close to major rivers, with NPD positively related with DMR, indicating the role of high-recharge major aquifers in facilitating the decoupling of the historical proximity of human settlements to rivers.” Supplementary Table 1. Slope of NPD vs DMR in 2010 for each aquifer type Aquifer types Groundwater zone
Area Recharge percentage rate (%) (mm/a)
slope
Major groundwater basins 11
